Glossary
Short definitions of Overshow and search terminology.
Last updated: 2 April 2026
Terms
OCR (optical character recognition). Extracting readable text from images of your screen so on-screen words become searchable.
Semantic search is not included in the shipping build, which uses full-text search. The term means retrieval that compares text representations for similarity.
Grounded answer. An answer retrieved from your own captured data, presented with source references so you can verify it.
MCP (Model Context Protocol). A standard for connecting AI assistants to external tools and data sources; Overshow can expose search and related tools through an MCP server.
STT / speech-to-text. Converting spoken audio into written text that can be indexed and searched.
Full-text search. Finding exact (or stemmed) keyword matches across captured text fields.
Exclusion categories. Groups of applications that Overshow can automatically exclude from capture according to policy or your settings.
Action prompts. Structured operations you can run against captured meeting-related data (for example summaries or follow-ups), defined by the product's meeting workflow.
Pre-meeting brief. A short summary of relevant past context prepared before a scheduled meeting so you arrive oriented.
Text embeddings are not used in shipping builds. They are numeric representations for similarity search, used only in development experiments.